CV-VRCam Depth of Field
Posted: 13 May 2016 04:50 PM   [ Ignore ]  
Total Posts:  6
Joined  2016-05-13

Is it possible to get this to work in any way? I was changing the physical properties on the camera rig to no effect before I remembered the rig is just for viewport editing and not for rendering.

The reason I wanted to try it out is that I noticed with 3D VR, when detailed elements at different depths overlap each other, the eye doesn’t know which one is supposed to be in front or behind creating more of a double vision effect.

This is most pronounced with title overlays, because of the negative spaces in the letter forms, as the eye scans over them, the background elements are still sharply in focus and is very dissonant.

Profile
 
 
Posted: 15 May 2016 06:35 PM   [ Ignore ]   [ # 1 ]  
Total Posts:  6
Joined  2016-05-13

Here are a couple examples of the “negative space” issue I am talking about, they need to be viewed in VR with Google cardboard or a VR viewer.

Try and read the credits in front of the desk:
http://www.youtube.com/watch?v=qECnb4CT-9c&t=4m4s

On this one try to focus on the light right in the center of the “O” in “ZERO”:
https://www.youtube.com/watch?v=VuRTTKP9vnM

Here’s an interesting discussion on the subtle cues that occur in real life when one object occludes another:
http://physics.stackexchange.com/questions/111006/how-does-light-bend-around-my-finger-tip

Profile
 
 
Posted: 24 May 2016 05:14 AM   [ Ignore ]   [ # 2 ]  
Administrator
Avatar
Total Posts:  406
Joined  2006-01-24

My understanding is that depth-of-field usually doesn’t work so well in VR because you can’t adapt it to the user’s focus without eye tracking in the headset.

That said, the depth pass renders something that looks like it could work, but you’d need a VR-aware blur option in composting (so that it doesn’t blur across the seams. I’m not sure if that’ll work or not.

Profile
 
 
Posted: 01 June 2016 03:46 PM   [ Ignore ]   [ # 3 ]  
Total Posts:  6
Joined  2016-05-13

Just a follow-up, I tried out this test video:
https://www.youtube.com/watch?v=NtOBbk_x3SM

It’s a bit of a mash-up of a few things including rotating the CV-VRCam rig and trying different eye distance settings, but if you notice the text, the “e” is hard to read in the lower right because the background is static and in focus, but when the background is moving or blurred the text becomes a lot easier to read.

Depth of field was just a thought, in reality I would probably avoid it anyways because of render times. But blurring the texture, or creating movement seems like an easy enough solution as well.

Profile